Multipath routing in wireless mesh networks

摘要: Multipath Routing in Wireless Mesh Networks Marc Mosko ∗ Palo Alto Research Center 3333 Coyote Hill Road Alto, CA 94304 Email: mmosko@parc.com Abstract— This paper addresses multipath routing a mobile wireless network. We review the premise that protocol should prefer disjoint path construction and argue using paths limits route reliability ad hoc networks compared to multiple loop-free need not be disjoint. In network, link lifetimes may relatively short traffic flows. The characteristics of MANET are significantly different than considered by Kleinrock his original delay analysis alternate routing. particular, on-demand protocols suffer significant during discovery. exploit mesh connectivity over non-disjoint improve s, t-connectedness lifetime Exploiting amortizes expensive discovery operations lead better performance or maximally paths. I. I NTRODUCTION main objective network is use several good reach destinations, just one best [1], without imposing excessive control overhead maintaining such has long been recognized as an important feature adapt load increase [2], [3]. Telecommunication adopted rout- ing, really form failover, 1984 [4]. Many papers on suggest proposed operate correctly (i.e., provide loop- free paths), specifically addressing when multipaths 1 used [5]–[9]. Other building paths, but claims correct operation (e.g. [10]–[13]). Several measure coupling [14]–[16], mutual interference routes common-channel multi-hop find with low coupling. Route coupling, however, makes every flow dependent other through area do address cost low- coupled protocol; they typically link-state pro-active protocols. Most works restrict number potential small number, usually two. AOMDV [17] allows up k link-disjoint RREPs, where “quickest” others chosen from next RREQs. SMR [18] builds two quickest RREQ then collects RREQs term ”multipath” denote set destination node edge J.J. Garcia-Luna-Aceves ∗† Computer Engineering Department University California at Santa Cruz Cruz, 95064 jj@soe.ucsc.edu for period chooses second first. zone-disjoint scheme [16], only built, necessarily minimum. uses iterative algorithm discard worst choice each round until left. this paper, we fully rich packet delivery. nutshell, well-designed many destinations keep those alive sending some amount data them function their quality. Paths poor quality longer distance used. exact methods propagate metrics distribute between open ques- tion. Interestingly, attempt take advantage advocate node- edge-disjoint Section II surveys literature case necessary networks. Furthermore, III shows well-connected offer substantially sets Based these results, IV illustrates approach which enforced, DOS [19] example. V summarizes implementation simulation study presented VI, compares distributions our can maintain 1.2 1.5 per hop, any special maintenance mechanisms. 100-node simulations, about 1/3 min-hop slightly higher delivery ratio. II. P RIOR W ORK literature, there types 2 , no common nodes except first (source) last (destination). edges, though nodes. = {s, a, b, c, t} m, n, although share b. There also zone try separated hops. Two “maximally”